This is a repostiory inteded for DCAE2022 Task 1: Low-Complexity Acoustic Scene Classification. The log-mel features can be downloaded in a numpy file format from this link: https://drive.google.com/drive/folders/1227rrHGJ6GBj63FwWTx3r-qK4yILm1pZ?usp=sharing
The link to pruning procedure is : https://github.com/Arshdeep-Singh-Boparai/passive-filter-pruning-Interspeech22
The script can be used to produce results (accuracy, log-loss) corresponding to the quantized models.
Confusion matrices produced using the submitted systems (4) for DCASE 2022 task1 development validation dataset.
keras, tensorflow, h5py, sklearn, scipy, dcase_utils...
More information will be updated soon.